Developer
Build and self-host
Install the MIT-licensed engine, record the bundled app, compile it, lint and certify the bundle, then replay it locally. No account or hosted service is required.
OpenAdapt.AI
LOCAL-FIRST · MIT OPEN SOURCE1.6k stars on OpenAdapt· 258 forks
Compile repeated GUI work into governed, deterministic workflows.
Demonstrate a bounded, repeated browser task. Compile it into a locally executable workflow whose healthy replay makes no model calls. When the interface changes, OpenAdapt re-resolves from retained evidence, proposes a governed repair, or stops for an operator.
Start with the supported browser path, then run through the local engine, managed control plane, or a qualified customer boundary. See how deployment works.
1.0click'Sign In'template12ms✓
2.0typeusername—8ms✓
3.0click'Tasks'template9ms✓
4.0click'Open referral'template11ms✓
5.0click'Save Encounter'⚠ UI drift detected
└ re-resolved via geometry — anchor update saved as diff
postconditions verified · run complete
total 2.4s · model calls: 0 · cost per run: $0.00VERIFIED · 0 MODEL CALLS
Demonstrate a bounded workflow. Replay deterministically and locally.
Healthy runs make zero model calls.
Drift is resolved, reviewed, or refused.
Choose local, hosted, or customer-controlled execution.
Choose your path
One engine, three launch paths
Automation & operations
Qualify a managed browser workflow
Bring one repeated browser workflow, its operating boundary, and an independent success oracle. If it fits the current managed path, scope a supervised pilot with the team.
Regulated enterprise
Deploy inside a trusted boundary
Define identity coverage, effect verification, artifact sanitization, and the runtime data boundary for a consequential workflow before it enters production.
Process
How it works
A demonstrated task becomes a reviewable program, not a prompt that a model reinterprets on every run.
- 1.0
Record
Demonstrate one bounded instance. OpenAdapt captures the browser evidence and input events needed to compile it.
Record — real footage · driving a live OpenEMR instance - 2.0
Compile
Your recording becomes a script you can read, edit, and reuse.
Compile — crafted annotation over a real recorded frame - 3.0
Replay
Healthy runs execute the compiled steps locally without a model call.
Run — real footage · driving a live OpenEMR instance - 4.0
Resolve or halt
Under drift, deterministic evidence re-finds the target, an optional model proposes a repair, or verification refuses to continue.
Governed drift handling — real footage · driving our MockMed demo app - 5.0
Verify and report
Configured postconditions and effects are checked, and every run records what happened, changed, or halted.
Audit — real footage · driving our MockMed demo app
UI drift
What “repair” means, and where it stops
OpenAdapt is not an unconstrained agent. Drift ends in one of four explicit outcomes, each visible in the run report.
Automatic
Deterministic re-resolution
Structure, templates, OCR, and geometry try to re-find the recorded target. No model is required. A changed anchor is saved as an auditable bundle diff.
Optional
AI-assisted proposal
If deterministic evidence is insufficient and a model is configured, it can propose a target or confirm limited visual state. The model can be wrong; identity, policy, and postcondition gates still decide whether execution continues.
Operator path
Human teaching and resume
A halt produces evidence for an operator. The CLI can teach a correction, and durable mode can checkpoint verified progress for authenticated approval and resume.
Unsupported drift
Refuse instead of improvise
If no configured path can establish the target or expected result, the workflow stops. OpenAdapt does not claim general adaptation to arbitrary application changes.
Availability and refusal depend on recorded evidence and configuration. Identity checks cover armed steps only; system-of-record verification requires declared effects and a configured verifier. Read the measured limits.
From demonstration to operation
One governed workflow, end to end
OpenAdapt supports repeated browser workflows from recording through compiled replay, verification, repair, and reporting. Choose where execution runs without changing the workflow model.
- 01
Capture the workflow
Demonstrate a bounded, repeated browser task, including the evidence OpenAdapt needs to identify targets and verify outcomes.
- 02
Compile and policy-check
Turn the demonstration into an inspectable program with parameters, target evidence, postconditions, and an explicit OpenAdapt certification policy.
- 03
Run through the governed gate
The fail-closed run gate rechecks the exact bundle and selected policy. Healthy replay makes no model calls and produces structured evidence.
- 04
Resolve change under governance
Re-resolve from retained evidence, propose a reviewable repair when needed, and stop when configured verification fails.
| Substrate | Availability | Evidence state | Current scope |
|---|---|---|---|
| Browser / Playwright | Public beta | Reference lifecycle accepted | The reference record, compile, policy-check, replay, refusal, and report lifecycle runs in CI. Each target application still requires workflow qualification. |
| Windows UIA | Partner qualification | Acceptance in progress | A typed, fail-closed candidate driver is under real Windows qualification with an independent database oracle. It is not in the hosted browser launch candidate. |
| Native macOS | Partner qualification | Acceptance in progress | A native exact-window candidate is under permissioned TextEdit qualification. Ambiguous or foreground-mismatched windows halt; broader application support is not established. |
| RDP | Partner qualification | Acceptance in progress | The network RDP candidate uses strict framebuffer leases, viewport-drift refusal, current-frame readiness, and an independent guest-file oracle. It is not in the hosted browser launch candidate. |
| Citrix | Design partner needed | No ICA/HDX evidence | The generic remote-window safety floor can begin qualification, but Citrix requires the customer’s actual ICA/HDX environment. RDP evidence is not treated as Citrix evidence. |
Choose the execution boundary
Local engine
Build, inspect, and run browser workflows on your own machine with the MIT-licensed engine.
View the engineManaged browser execution
Operate approved browser workflows through the hosted control plane with run history, reports, usage, and governed updates.
See hosted optionsCustomer-controlled deployment
Keep runtime data inside a qualified customer boundary when the workflow involves regulated data, private systems, or deployment-specific effect checks.
Review the security boundaryTechnical teams can review the engine limits and hosted architecture before qualifying a production workflow.
Measured evidence for repeated browser work
Repeated work should not pay an agent to rethink the same task.
7.6× faster
median workflow time than the computer-use agent
$0 model cost
on a healthy compiled replay
0 model calls
on a healthy compiled replay
In a bounded browser benchmark, both approaches completed every tested run. OpenAdapt's median compiled replay was 7.6× faster and incurred no model cost on healthy runs. This benchmark measures speed and model cost for one task; it is not a production reliability claim.
Safety
Configured safeguards refuse instead of guessing.
OpenAdapt can require identity and effect evidence before a consequential step, then halt for an operator when verification is ambiguous. That protection applies only where the workflow is configured to check it; a success screen alone is not proof that a write reached the right system or record.
Who it's for
Teams closing a UI-only last-mile gap
The strongest fit starts with structured inputs, existing business logic, high repeated volume, a browser step without a supported API, and an independent source of truth for the intended effect. OpenAdapt compiles that bounded UI-only gap into an inspectable program. Healthy replay makes no model calls; consequential workflows still require explicit identity, effect, policy, storage, and retention configuration.
Optional PII/PHI scrubbing covers configured sharing paths, not every raw artifact.
Automation teams & BPO operators
- High-volume repeated operations with structured inputs, established business logic, and a UI-only last-mile gap that can be checked against an independent effect source of truth.
RCM & vertical-software vendors
- Products with supported APIs for the core path and a bounded browser-only step at the edge, where volume and a separate system of record make verification practical.
Example: Healthcare workflow reference
Regulated enterprise operations
- Governed teams that already know the inputs and rules, repeat the task at material volume, and can verify the external effect independently of the GUI run.
Example: Lending operations reference
Qualify one workflow with us
We first check the substrate, verification boundary, and measurable outcome. If the workflow fits, we can scope a paid design-partner pilot in your environment.
Qualify a workflowRunning the same workflow hundreds of times on software you can't API into?
OpenAdapt is launching managed browser execution now, with the same deterministic compiler available under MIT for local and customer-controlled deployment.
Choose a launch path →Launch options
Run it yourself or launch with us
Run the MIT-licensed engine yourself, use managed browser execution, or qualify a customer-controlled deployment. Hosted prices are confirmed by Stripe before payment.
Open Source
FreeMIT
For builders and self-hosters who want it running on their own machines.
- Browser artifacts stay local unless you explicitly push
- Deterministic re-resolution with auditable diffs
- Policy certification, fail-closed run gate, and reports
- MIT-licensed engine and CLI
pip install openadapt-flow
Managed browser
Hosted
Hosted execution
Bring an approved browser workflow to a managed control plane for repeat execution, reporting, and governed updates.
- Managed execution of approved browser workflows
- Deterministic healthy replay with zero model calls
- Run history, failure reports, usage, and governed updates
- Sanitized uploads admitted under declared policy
- Price and billing period confirmed in Stripe
Hosted upload accepts an approved sanitized copy, not the original recording. Live screens can contain sensitive data again; workflows that expose PHI require a separately qualified customer-controlled boundary. Review the security boundary.
Start with our teamBy subscribing, you agree to the Terms of Service and acknowledge the Privacy Notice.
Regulated deployment
Enterprise
Contact sales
For consequential workflows that require a customer-controlled data boundary. We qualify the substrate, artifact policy, effect oracle, and operating model before production use.
- Qualify the substrate and safety boundary first
- Scope one workflow and one measurable outcome
- Execute in the customer-controlled environment
- Identity and effect checks configured per deployment
- Sanitized derivatives may cross approved boundaries
- Runtime PHI remains in the trusted execution environment
- Deployment, support, and compliance terms documented in scope
Read our security posture: on-prem data handling, what is and isn't certified yet, and how to report a vulnerability.
Plan a regulated deploymentExecution boundary
Choose the operating model that fits the workflow
Use managed browser execution for approved public web workflows. Choose a customer-controlled deployment when runtime data, private systems, or effect verification must remain inside your boundary.
Managed browser subscriptions open only after the live checkout and account-return path pass launch qualification. Regulated deployment and service terms are scoped separately.
Open source
MIT licensed. Install it and read the code.
openadapt-flow is the canonical engine. Its browser path records a workflow, compiles it into a reviewable program, and replays it locally with no model calls on a healthy run.
Install the browser workflow CLI
One command installs the OpenAdapt launcher and governed compiler. No account or hosted service is required.
Install
$curl -fsSL https://openadapt.ai/install.sh | sh
The end-to-end browser loop
Try the whole loop right now against the bundled demo app: record, compile, inspect, certify, replay, and induce known test drift. No account or cloud service. Once it is installed, your workflow data stays on your machine. One heads-up on first run: the first demo-record or replay downloads a bundled Chromium (~150 MB) once, so that step takes a few minutes, and on Linux you may need playwright install-deps first.
curl -fsSL https://openadapt.ai/install.sh | shInstall OpenAdaptopenadapt flow demo-record --out recRecord yourself doing the taskopenadapt flow compile rec --out bundle --name my-taskTurn the recording into a workflowopenadapt flow lint bundleInspect identity, assertion, and risk coverage gapsopenadapt flow certify bundle --policy clinical-writeEnforce a policy before deploymentopenadapt flow replay bundleReplay it locally, with zero AI callsopenadapt flow replay bundle --drift themeExercise deterministic re-resolution on bundled test driftuv tool uninstall openadaptRemove the launcher when you are finishedEvery replay writes a step-by-step run report: what ran, what it saw, what re-resolved, and what halted. The browser path is the only shipped end-to-end backend today. Source and measured limits on openadapt-flow. Evaluating a regulated workflow? Book a demo.
PyPI Download Trends
Historical download statistics for OpenAdapt packages
8Packages
View:
Range:
Loading statistics...
Data sourced from PyPI OpenAdapt packages via pypistats.org API (data retained for ~180 days)
FAQ
Frequently asked questions
- What is OpenAdapt?
- OpenAdapt is an open-source governed compiler for repeated GUI work. You demonstrate a task, it compiles the trace into a deterministic local program, and configured verification decides whether UI drift can be re-resolved, needs a proposed repair, or must halt. A healthy replay makes no model calls.
- How is OpenAdapt different from RPA tools like UiPath?
- Conventional RPA is broad and mature, especially on Windows, but workflows are usually assembled around selectors and integrations. OpenAdapt targets the narrower case where work is repetitive and consequential but the interface is visual, variable, or integration-hostile. It compiles a demonstration and uses a resolution ladder under drift rather than asking a model to re-plan every run.
- How is OpenAdapt different from AI computer-use agents?
- A computer-use agent is appropriate for novel work because it can re-plan each run. OpenAdapt is for repeated work: healthy replay follows the compiled program locally with zero model calls. Under drift, deterministic structure, template, OCR, and geometry evidence run first. An optional model can propose a repair, but it does not make unsupported drift safe and remains subject to configured verification and policy.
- Does “repair” mean every UI change is handled automatically?
- No. A target may be re-resolved deterministically, an optional model may propose a repair, an operator may teach a correction, or the run may halt as unsupported. Available evidence, identity coverage, postconditions, policy, and verifier configuration determine which outcome is allowed. OpenAdapt does not claim general adaptation to arbitrary application changes.
- Does my data leave my machines?
- Self-hosted runs remain local by default. Hosted upload accepts an approved sanitized copy, not the original recording and not an assumption that compilation removed sensitive data. Live screenshots can contain sensitive data again, so they stay inside the declared managed, customer-controlled, or on-prem runtime boundary. See the security page for the exact admission and review controls.
- Is OpenAdapt free?
- The engine is MIT-licensed and free to use or modify. When production checkout is qualified and explicitly enabled, hosted browser execution is a paid subscription whose price and billing period are confirmed in Stripe. Regulated and customer-controlled deployments are scoped separately because the substrate, data boundary, verifier, and operating responsibilities differ.
- Is managed hosted execution available?
- Managed execution is scoped to approved browser workflows. The hosted control plane manages runs, reports, usage, and governed workflow updates. Direct subscription activation remains gated until the live checkout and account-return path complete launch qualification; workflow qualification is available through the team.
- What software does OpenAdapt work with?
- The supported launch path is repeated work in browser-based software. Workflows involving private systems, regulated runtime data, or deployment-specific effect checks can be qualified for a customer-controlled boundary. Technical evaluators can review the published engine limits before choosing a workflow.
Workflow qualification
Book a 30-minute automation fit call
Tell us the substrate, repetition, consequence of error, data boundary, and current verification path. We'll tell you whether it fits managed browser execution, a customer-controlled deployment, or neither.
Product updates and hosted launch information. This does not create an account or promise access.